in this project i created an obstacle avoiding robot using the q learnig algorithm which is based on learning from trial and error and giving rewards for the right move and penalties for the wrong move ,the same way humans learn
PROBLEM the object is to teach the robot to learn how to avoid obstacles after a certain number of episodes and from this dataset it can then the optimal action to take .
MATERIALS USED
1.ultrasonic sensor The SEN136B5B is an ultrasonic range finder from Seeedstudio. It detects the distance of the closest object in front of the sensor (from 3 cm up to 400 cm). It works by sending out a burst of ultrasound and listening for the echo when it bounces off of an object. It pings the obstacles with ultrasound. The Arduino board sends a short pulse to trigger the detection, then listens for a pulse on the same pin using the pulseIn() function. The duration of this second pulse is equal to the time taken by the ultrasound to travel to the object and back to the sensor. Using the speed of sound, this time can be converted to distance. [source :https://docs.arduino.cc/built-in-examples/sensors/Ping]
2.L298 H bridge motor driver
This module is based on the very popular L298 Double H-Bridge integrated circuit driver. The circuit will allow you to easily and independently control two motors up to 2A each in both directions.
It is ideal for robotics applications and well suited for connection to a microcontroller requiring only a few control lines per motor. It can also be interfaced with simple manual switches, TTL logic gates, relays, etc.
Characteristic
Motor power supply: 7 to 24 VDC
Control logic: Standard TTL level logic
Output Power: Up to 2A, 25W each
Enable and Control Direction Pins
Radiator for IC
LED power indicator
4 Direction LED Indicators
3.powerbank
4.jumper wires
5.dc motors
- Two wheel robot car chasis
7.arduino uno
Code Explanation
int ACTIONS = [0,1,2,3]
//deines the actions the robot must perform which are forward backward left and stop //with 0 forward 1 backward 2 left and 3 right
int NEXT_STATE; int STATE = 0; NEXT_STATE = STATE+1;
//this is the state that the robot moves into after performing the first action //the initial state of the robot in the beginning of a episode is survive
float Q[N_STATES][NUMBER_OF_ACTIONS] = {{0.0,0.0,0.0,0.0},
{0.0,0.0,0.0,0.0},
{0.0,0.0,0.0,0.0},
{0.0,0.0,0.0,0.0},
{0.0,0.0,0.0,0.0},
{0.0,0.0,0.0,0.0},
{0.0,0.0,0.0,0.0},
{0.0,0.0,0.0,0.0},
{0.0,0.0,0.0,0.0},
{0.0,0.0,0.0,0.0}};
//